"#$%&'() $(* +,'() -($.

/)0 '( 1$2

3$( 45(6#7

And I cherish more than anything else the Analogies, my most
trustworthy masters. 1hey know all the secrets o Nature, and they
ought to be least neglected.
- Johannes Kepler

89 8(67/*5%6'/(

Analogy plays a central role in legal reasoning, yet how to analogize is
poorly taught and poorly practiced. \e all recognize when legal analogies are
being made: when a law proessor suggests a diicult hypothetical in class and
a student tentatiely guesses at the answer based on the cases she read the
night beore, when an attorney adises a client to settle because a preious
case goes against him, or when a judge adopts one precedent oer another on
the basis that it better its the present case. loweer, when it comes to
explaining why certain analogies are compelling, persuasie, or better than the
alternatie, lawyers usually draw a blank. 1hey hae little idea how to create
an analogy, what an analogy is, or why one analogy might be more eectie
than any other. 1he teaching o analogy reinorces this sense that analogies
are a mystery: the teacher suggests that the student will learn what is a good
analogy only through experience. I law proessors try to teach analogy
construction at all, they usually choose one o two approaches. 1hey might
teach the LSA1 Model` o analogies: Arm is to Leg as land is to \hat`
1he student should answer loot,` but een so, what has the student learned
At best, this model suggests that an analogy is some kind o relationship, but
it doesn`t gie much guidance beyond that. Alternatiely, the proessor might
adopt the theories o a number o legal philosophers
1
and teach analogies as a
type o rule. In this model, the student takes a case, renders the case down
into a holding, and then applies the holding to the undecided case. 1his
model has the beneit o teaching the student that analogies actually matter in
law, but it suggests that the only trick o legal analogies is to ind the
appropriate rule that explains a case. \et, as students quickly realize, cases

Robert l. Irwin IV 1erm Assistant Proessor o Legal Studies, \harton School,

Uniersity o Pennsylania ,hunterdwharton.upenn.edu,. 1his is an edited ersion o an
earlier article, Dan lunter, "#$%&' (% )&& *$+,#- .'$/&,0 $'1 2+#3#1#'4 (' *$5, 50 Lmory L.J. 119
,2001,. Used with the permission o Lmory Law Journal.
1. 6## lunter, biographical ootnote %78+$, at 1239-43.
]ovrvat of tbe ...ociatiov of egat !ritivg Director. !"#$% '

()'
*+,,#- ./ /+01$2 .#1$/3 3#4, -# + 0156$/ 78$/9 +,3 -:/2 #;-/, 5+2 ./ 80/3 -#
0866#7- <81-/ *#,;$1*-1,= 671,*16$/0%
>,/ #; -:/ 5+1, 67#.$/50 41-: .#-: -:/0/ 5#3/$0
'
10 -:+-9 4:1$/ -:/2
*#,-+1, + =7+1, #; -78-:9 -:/2 +7/ 0# =/,/7+$ +0 -# ./ #; ,# 67+*-1*+$ 80/ 1,
/?6$+1,1,= -# 0-83/,-0 #7 $+42/70 :#4 -# 80/ +,+$#=1/0 /;;/*-1@/$2% A- vv.t ./
-:/ *+0/ -:+- 0#5/ +,+$#=1/0 +7/ ./--/7 -:+, #-:/709 .8- ,/1-:/7 #; -:/0/
5#3/$0 /?6$+1,0 4:2 0#5/ +7/ ./--/7 -:+, #-:/709 #7 :#4 + 0-83/,- #7 + $+42/7
*+, $/+7, -# 310-1,=810: =##3 +,+$#=1/0 ;7#5 5/31#*7/ #,/0%
B:/ 6876#0/ #; -:10 +7-1*$/ 10 -# 67#@13/ + 0156$/ 5#3/$ -:+- *+, ./ 80/3
-# -/+*: +,3 -# $/+7, :#4 +,+$#=2 +*-8+$$2 4#7C09 +,3 4:+- 5+C/0 #,/ +,+$#=2
086/71#7 -# + *#56/-1,= +,+$#=2% B:/ 5#3/$ 10 37+4, ;7#5 + ,85./7 #;
-:/#71/0 #; +,+$#=2D5+C1,= 1, *#=,1-1@/ 0*1/,*/% E#=,1-1@/ 0*1/,*/ 10 -:/
F$#,=D-/75 /,-/76710/ -# 8,3/70-+,3 -:/ 51,3 0*1/,-1;1*+$$2%G
H
B:/ ;1/$3 0-831/0
-:/ 5/*:+,1050 -:+- +7/ 1,@#$@/3 1, *#=,1-1@/ 67#*/00/0 $1C/ -:1,C1,=9
5/5#729 $/+7,1,=9 +,3 7/*+$$I +,3 #,/ #; 1-0 5+1, ;#*1 :+0 .//, #, :#4 6/#6$/
*#,0-78*- +,+$#=1/0% B:/ $/00#,0 ;7#5 *#=,1-1@/ 0*1/,*/ -:/#71/0 #; +,+$#=2
*+, ./ +66$1/3 -# $/=+$ +,+$#=1/0 -# =1@/ 0-83/,-0 +,3 $+42/70 + ./--/7
8,3/70-+,31,= #; -:10 ;8,3+5/,-+$ 67#*/00 1, $/=+$ 7/+0#,1,=%

!!" $%& '()*+,+-& ./+&*/& (0 1*23()4

B:/7/ 10 ,# 4#73 4:1*: 10 80/3 5#7/ $##0/$29 #7 1, + =7/+-/7 @+71/-2 #;
0/,0/09 -:+, J,+$#=2%
D K#:, L-8+7- M1$$

B:/ -/75 F+,+$#=2G 10 80/3 1, 0# 5+,2 31;;/7/,- 4+209 /06/*1+$$2 41-:1,
$+49 -:+- 1- 10 31;;1*8$- -# 3/;1,/ 1- +**87+-/$29 $/- +$#,/ *#567/:/,01@/$2%
N87-:/75#7/9 $/=+$ *#55/,-+-#70 :+@/ *+80/3 /,#75#80 67#.$/50 .2 ;+1$1,=
-# /?6$+1, :#4 +,+$#=2 31;;/70 ;7#5 -:/ 7/$+-/3 1,;/7/,*/ 67#*/00/0 #;
1,38*-1#, +,3 5/-+6:#7% B:10 :+0 $/3 -# 0$#662 -:1,C1,=9 +,3 6##7 +,+$2010%
A- 10 156#7-+,- -# 3/;1,/ +,+$#=2 +,3 +,+$#=1*+$ 7/+0#,1,=9 +,3 -# 0:#4 :#4
-:/2 +7/ 31;;/7/,- ;7#5 .#-: 1,38*-1@/ 1,;/7/,*/ +,3 5/-+6:#7%
J, +,+$#=2 10 + ,#,D13/,-1*+$ #7 ,#,D$1-/7+$ 0151$+71-2 *#56+710#, ./-4//,
-4# -:1,=09 41-: + 7/08$-1,= 67/31*-1@/ #7 /?6$+,+-#72 /;;/*-%
O
B:10 5/+,0 -:+-
-4# 1-/50 +7/ *#56+7/39 +,3 -:/ #8-*#5/9 7/08$- #7 3/-/751,+-1#, #; #,/ #;
-:/0/ 1-/50 10 67/31*-/3 #7 /?6$+1,/3 -# ./ -:/ 0+5/ +0 -:/ #-:/7% B261*+$$29
4/ *#56+7/ 1-/50 4:/7/ 4/ C,#4 -:/ #8-*#5/ #; #,/9 +,3 4/ 08==/0- -:+-
-:/ #8-*#5/ #; -:/ #-:/7 41$$ ./ -:/ 0+5/%

'% B:/ #-:/7 5+1, 67#.$/5 10 -:+- -:/0/ -:/#71/0 +7/ +$0# 47#,=% ee ia. +- ('OHDO)%
H% P+@13 Q% R7//, /- +$%9 Cogvitire cievce - .v vtroavctiov ' ST$+*C4/$$ U8.$10:/70
(VVWX%
O% B:10 3/;1,1-1#, 10 +3+6-/3 ;7#5 P/37/ R/,-,/7 /- +$%9 1ierivg Metabor a. .vatog,9 1,
.vatogicat Rea.ovivg: Per.ectire. of .rtificiat vtettigevce, Cogvitire cievce, ava Pbito.ob, (Y( SP+,1/$ Z%
Z/$5+, /3%9 [$84/7 (V\\X !:/7/1,+;-/7 .vatogicat Rea.ovivg]%
lall 2004| !"#$%&'( #'* +,&'( -'#./(0 &' 1#2

153
1he obious example o legal analogy is case law reasoning: prior analogs,
called precedents, are used to predict, explain or justiy the outcome o the
currently undecided case.
5
An example will be useul here to explain the
mechanism. Let us say a judge is aced with a case in which a man is suing or
the loss o his luggage while he was traeling onboard an oernight erry. 1he
luggage was stolen rom an oerhead rack in the plainti`s compartment. 1he
plainti is suing the erry company on the theory that it was the bailee o his
bags and thereore liable or their loss. Let us assume that there is no
statutory pronouncement on the subject, and there are only two precedents
that might be releant to the decision. 1he irst precedent inoled a hotel
proprietor who was ound liable or a guest`s stolen luggage because part o
the contract o hospitality inoled reasonably sae storage o the guest`s
belongings. 1he second precedent inoled a railroad company, which was
ound not liable or the loss o the luggage o a passenger who traeled in a
sleeper berth because the contract was primarily or trael and not or
lodging.
6

Plainti`s counsel will suggest that the erry is a loating hotel`: it has
cabins like hotel rooms, it has restaurants and other hotel-like acilities, and
passengers almost ineitably spend the night onboard. 1hereore, the judge
should ollow the hotel precedent and ind or Plainti. Deendant`s counsel
will suggest that the erry is a seagoing train`: the intention o the passengers
is primarily to trael and not to stay oernight, trains hae restaurants and
bars, and so orth. lence, the decision should be against Plainti.
1he question or the judge is straightorward: \ill she opt or the
loating hotel` analogy or the seagoing train` analogy In deciding this
case, the judge may draw an analogy to a hotelkeeper`s duty to protect the
bags o a guest, or alternatiely, point to the railroad`s lack o duty to do the
same or a passenger. In each analogy, the judge chooses some non-identical
eatures o the precedent as being suiciently similar to the current case to
warrant the same outcome - perhaps the act that both the train and erry
inole transportation or alternatiely that both the hotel and erry hae
sleeping quarters. laing drawn the comparison between the precedent and
the undecided erry case, the judge uses the precedent`s outcome to decide the
new case. \hether the judge chooses the train or hotel precedent as
analogous, the analogy has a strongly constraining, predictie eect on the
outcome o the undecided erry case.
Inductie inerence or induction is related to analogy. Induction is,
generally, the process o taking a number o speciic cases or instances,
classiying them into categories according to releant attributes and outcomes,

5. Ldward Lei, -' 3'45/*6$4&/' 4/ 1"(#. 7"#,/'&'( 9-15 ,U. o Chi. Press 1949,8
6. 1his 9"*#':"' experiment is based on the case o -*#;, . <"2 ="5,"0 >4"#;?/#4 @/8,
151 N.\. 163 ,1896,. 1hanks go to Martin P. Golding, 1"(#. 7"#,/'&'( ,Knop 1984, or
uncoering this gem.
!"#$%&' ") *+, -.."/0&*0"% ") 1,2&' 3$0*0%2 40$,/*"$. |Vol. 2

154
and generalizing an inclusory rule rom them.

1hat is, we take a number o

isolated experiences and attempt to explain them by a general rule that coers
the instances examined. \ithin law, we can ind numerous examples o this:
the ability o attorneys to say rom experience what particular judges`
decisions are likely to be, their ability to it multiple precedents into a coherent
ramework o rules and principles, and so on.
8

Many lawyers ail to recognize that induction requires the generalization
o a rule rom prior experience, whereas analogy is a one-to-one similarity
comparison that requires no generalization to operate eectiely. 1he best
example o this is the ,5#.6,7 2,%,$0. canon o statutory interpretation.
9
85#.6,7
2,%,$0. means o the same kind, genus or nature`
10
and the canon is inoked
when there is a statutory deinition o the orm 9, :, ; or other.`
11
1he
canon is applied to determine whether something that is not deined should
be included as other.` 1hat is, it is used to deine the scope o general words
that immediately ollow speciic words. So, or example, in a statutory
deinition that indicated planes, trains, automobiles, or other,` a question
might arise whether a skateboard, a solar-powered ehicle, or a snowmobile
should be included within the deinition o other.`
1he ,5#.6,7 2,%,$0. rule clearly relies on inductie inerence. 1o determine
what the expression other` encompasses, courts must examine the words
that make up the antecedent part o the expression in question, and then
decide what the scope o those words incorporates. Courts speciically
require the identiication o the genus as the irst step in the interpretie
process. It will thereore always be necessary to create an inductie
generalization about the set o concepts that the legislature intended to be
included.
12


. <,, John l. lolland et al., =%6#/*0"%> ?$"/,..,. ") =%),$,%/,@ 1,&$%0%2@ &%6 40./"A,$: ,MI1
Press 1986,, Keith J. lolyoak & Paul 1hagard, B,%*&' 1,&C.> -%&'"2: 0% D$,&*0A, E+"#2+* 19-20
,MI1 Press 1995,, L. Jonathan Cohen, -% =%*$"6#/*0"% *" *+, ?+0'"."C+: ") =%6#/*0"% &%6 ?$"F&F0'0*:
,Oxord U. Press 1989,, Nicholas Rescher, ?,0$/,G. ?+0'"."C+: ") </0,%/,> D$0*0/&' <*#60,. 0% H0.
E+,"$: ") =%6#/*0"% &%6 </0,%*0)0/ B,*+"6 ,U. o Notre Dame Press 198,, Nicholas Rescher,
=%6#/*0"%> -% 8..&: "% *+, !#.*0)0/&*0"% ") =%6#/*0A, I,&."%0%2 ,Blackwell 1980,. lor an analysis o the
similarities between analogy and induction, see Stuart J. Russell, -%&'"20/&' &%6 =%6#/*0A, I,&."%0%2
,unpublished Ph.D. dissertation, Stanord U. 199, ,microormed on 1echnical Rpt., Dept. o
Computer Sci., S1AN-CS-8-1150,, Stuart J. Russell, E+, J., ") K%"L',62, 0% -%&'"2: &%6 =%6#/*0"%
,Morgan Kauman 1989,.
8. lor an analysis o inductie inerence in law, see Dan lunter, M" 30'6,$%,.. ") <0%2',
=%.*&%/,.> =%6#/*0A, =%),$,%/, 0% 1&L, 48 J. Leg. Lduc. 365-401 ,1998,, Michael S. Moore, ?$,/,6,%*@
=%6#/*0"%@ &%6 8*+0/&' N,%,$&'0;&*0"%, in ?$,/,6,%* 0% 1&L 183 ,Laurence Goldstein ed., Oxord U.
Press 198,.
9. 1he rule applies also in interpretation o other documents, but is at its most powerul
in statutory interpretation. <,, 2,%,$&'': J. Bell & G. Lngle, D$".. "% <*&*#*"$: =%*,$C$,*&*0"% 135-3
,3d ed., Butterworths 1995,, l.A.R. Bennion, <*&*#*"$: =%*,$C$,*&*0"%> - D"6, 858-64 ,2d ed.,
Butterworths 1992,.
10. Roger Bird, O.F"$%G. D"%/0., 1&L 40/*0"%&$: ,th ed., Sweet & Maxwell 1983,.
11. Under some conditions, the ,5#.6,7 2,%,$0. canon will be applied een where there is
no speciic reerence to other.` <,, Bell & Lngle, .#C$& n. 9, at 135-3.
12. Scott Brewer wrongly suggests that the rule is analogical. Scott Brewer, 89,7C'&$:
I,&."%0%2> <,7&%*0/.@ ?$&27&*0/.@ &%6 *+, I&*0"%&' =%*,2$0*: ") 1,2&' -%&'"2: 93 ,U. o 1oronto 1995,.
lall 2004| 1eacbivg ava |.ivg .vatog, iv ar

155
Analogy is dierent rom induction. It does not rely on any
generalization o prior experience. loweer, it is related to induction in that
both rely on similarity comparisons o prior experience.
Analogical inerence also shares a number o eatures with metaphorical
inerence. Metaphors hae a less constraining eect on reasoning than do
analogies, but they operate in a similar ashion. Metaphor is an expression
orming a non-literal similarity comparison between two things, which has an
expressie or aectie content and thereby carries meaning.
13
Unlike
analogies, metaphors do not hae a predictie content and do not strongly
constrain the outcome o the reasoning process.
14
1hus, in Shakespeare`s
Roveo ava ]vtiet, Romeo`s metaphor Juliet is the sun` contains an explicit
similarity comparison between Juliet` and the sun`: in comparing Juliet and
the sun, the intent is to coney the expressie meaning that, to Romeo, she is
the light o his world, she proides lie and energy. Metaphors within the
legal sphere operate in a similar way. 1o use the metaphor o the corporate
eil` - to say or example that Company directors are protected by the
corporate eil` - coneys the expressie meanings that the corporation is a
separate entity rom the directors, the directors are shielded behind some kind
o protection, the directors are somehow unseen` while behind the eil, and
so on.
15
loweer, the metaphor does not strongly direct or predict the
outcome o a case in the way that a legal analogy does. In contrast, or
example, i our judge decides that a erry is just the same as a train, the train`
precedent controls the outcome o the undecided erry case.
1he deinitions gien aboe are deried rom cognitie science
approaches to metaphor and analogy,
16
and they are useul in both deeloping
the connection between the two concepts and explaining their dierences.
Both analogy and metaphor inole a similarity relation between two objects,
and the similarity relation transers meaning rom one object ,the source, to
another ,the target,.
1
1he major dierence between the two is that an
analogy has an explicit explanatory or predictie component which metaphors
lack.

lor a longer discussion o the e;v.aev geveri. rule and its inductie character, see lunter, .vra n.
8.
13. Again adapted rom Gentner et al., .vra n. 4, at 11.
14. a.
15. 1his example stems rom the general metaphor o the 1he Corporation is a
Person,` a metaphor analyzed by Sanord A. Schane, 1be Cororatiov i. a Per.ov: 1be avgvage of a
egat ictiov, 61 1ul. L. Re. 563 ,198,. Schane`s exemplary study demonstrates that the
metaphor is not merely a iction, created out o legal whole cloth.` Instead, the metaphor
accords with generally held conceptions that humans hae about institutions. 1hrough
linguistic examples, he traces how we reer to institutions made up o a number o indiiduals
as though they were a single unit or person. a. at 595. lor an earlier ormulation o a similar
proposition, see A.\. Machen, Jr., Cororate Per.ovatit,, 24 lar. L. Re. 253, 263 ,1911,.
16. Gentner et al., .vra n. 4, at 11.
1. Bipin Indurkhya, Metabor ava Cogvitiov: .v vteractiovi.t .roacb 14-1 ,Kluwer
1992,, Gentner et al., .vra n. 4.
]ovrvat of tbe ...ociatiov of egat !ritivg Director. !"#$% '

()*
+,-. /01$2324-#3 -. 3#4 -34/35/5 4# 264-78$24/ 7#91$/4/$: 4,/ 6/$24-#3.,-1
;/4<//3 9/421,#6 235 232$#=:%
(>
?3.4/25 -4 -. #@@/6/5 4# .,#< 4,24 ;#4, 26/
.-9-$26-4:A9247,-3= 16#7/../. <,-7, -346#587/ 2 7#3.462-34 81#3 ,8923
4,-3B-3=C 235 9#6/#D/6 4# 16#D-5/ .#9/ =8-5237/ 2;#84 4,/ 5-@@/6/37/
;/4<//3 E232$#=-/.CF E9/421,#6.CF 235 E-35874-#3%F G-4, 4,/./ 5/@-3-4-#3.
#84 #@ 4,/ <2:C $/4 8. $##B 24 ,#< ,8923. 92B/ 232$#=-/.%

H% H32$#=: 4,/#6-/. -3 7#=3-4-D/ .7-/37/

H35 6/9/9;/6C 5# 3#4 3/=$/74 D2=8/ 232$#=-/.% I84 -@ :#8 <-., 4,/9
6/.1/742;$/C 46: 4# 7$26-@: 4,/9%
A J/#6=/ K#$:2

H32$#=: ,2. ;//3 4,/ .8;L/74 #@ 2 .-=3-@-7234 ;#5: #@ 6/./267, <-4,-3
7#=3-4-D/ .7-/37/C <,/6/ 987, #@ 4,/ /91,2.-. ,2. ;//3 #3 ,#< ,8923.
4,-3B 76/24-D/$:% M#.4 6/./267,/6. 2=6// 4,24 232$#=: -. @83529/342$ N #6 24
$/2.4 D/6: .-=3-@-7234 N -3 76/24-D/ 4,#8=,4%
(O
+,#8=, 4,/6/ 26/ D26-#8.
7#91/4-3= 9#5/$. 4,24 5-@@/6 -3 4,/-6 9-384-2/
'P
7#=3-4-D/ .7-/34-.4. =/3/62$$:
2=6// #3 4,/ @83529/342$.% +,/ 92-3 @/2486/. 26/ 4,/ 7#37/14 #@ 9211-3=C
4,/ 6#$/ #@ .#867/ 235 426=/4 5#92-3.C 235 4,/ #1/624-#3 #@ 4,/ 1264. #@ 4,/
16#7/..%
I#4, 232$#=-72$ 235 9/421,#6-72$ 6/2.#3-3= -3D#$D/ 2 9211-3= #@
7#37/14. @6#9 #3/ ./4 #@ -5/2. Q4,/ .#867/ 5#92-3R 4# 23#4,/6 ./4 #@ -5/2.
Q4,/ 426=/4 5#92-3R% +,/ 2116#27, 723 ;/ 5/9#3.4624/5 ;: .//-3= ,#< -4 -.
8./5 4# /01$2-3 9/421,#6% 1arget 235 .ovrce 26/ .#9/4-9/. 5/.76-;/5 2. 4,/
E4<# ,2$D/. #@ 9/421,#6%F
'(
?3 23: =-D/3 9/421,#6C 4,/ 426=/4 -. 4,/

(>% S#6 9#6/ 5/42-$/5 /029-324-#3 #@ 4,-. 1#-34C .// Q29#3=.4 #4,/6.R T/56/ J/343/6
235 M% U/V-#6.B-C 1be bift frov Metabor to .vatog, iv !e.terv cievceC -3 Metabor ava 1bovgbt WWX
QH356/< Y64#3: /5%C '5 /5%C Z29;6-5=/ [% K6/.. (OO\R] ?3586B,:2C .vra 3% (XC 24 \()A)*] ^26$
_% M27Z#6927C . Cogvitire 1beor, of Metabor '\ QM?+ K6/.. (O>)R] M26B U#,3.#3C ove
Cov.traivt. ov vboaiea .vatogicat |vaer.tavaivgC -3 .vatogicat Rea.ovivgC .vra 3% W 24 ')%
(O% M%G% ^:./37B ` M%+% a/23/C Cogvitire P.,cbotog,: . tvaevt`. avaboo/ \O'AOW
Qb-$$.52$/ (OO)R] H% J263,29 ` U% Y2B,-$$C 1biv/ivg ava Rea.ovivg '()A(X QI$27B</$$ (OOWR%
'P% +,/ 9#.4 -34/6/.4-3= 235 4,#8=,4A16#D#B-3= #@ 2$$ 4,/#6-/. #@ 232$#=: 5-@@/6. -3 -4.
$26=/6 @/2486/.% +,-. -. 4,/ <#6B #@ T#8=$2. b#@.4254/6 235 ,-. .485/34.% ee e.g. T#8=$2. _
b#@.4254/6C tvia Covcet. ava Creatire .vatogie.: Covvter Moaet. of tbe vvaavevtat Mecbavi.v. of
1bovgbt QI2.-7 I##B. (OO)R] T#8=$2. _ b#@.4254/6 ` M/$23-/ M-47,/$$C .v Orerrier of tbe Co,cat
Pro;ect Q9-76#@#69/5 #3 +/7,3-72$ _14% Z_ZZA)'A(OO(C Z46% @#6 _/./267, #3 Z#37/14. 235
Z#=3-4-#3C ?35-232 [%C I$##9-3=4#3R] M/$23-/ M-47,/$$C .vatog,Ma/ivg a. Percetiov: . Covvter
Moaet QM?+ K6/.. (OO\R% b#@.4254/6 235 ,-. 4/29 8./ 7#918424-#32$ 9#5/$-3= #@ D/6: $#<A$/D/$
16#7/../. -3 232$#=: @6#9 4#: 5#92-3. .87, 2. $/44/6A.46-3=.C 389;/6 ./c8/37/. 235 2$1,2;/4
5/.-=3.% H3 /26$: 6/./267, 9#5/$ #@ ,-. <2. 25#14/5 -3 2 16#4#4:1/ 4/27,-3= 4##$ @#6 $2<
.485/34.% K/4/6 d8;/6C .vatog, erci.e. for 1eacbivg egat Rea.ovivgC (X U% e/=% d485% ` ^587% O(
Q(O>>R%
'(% ?%H% _-7,265.C 1be Pbito.ob, of Rbetoric O) QY0@#65 [% K6/.. (O\*R% _-7,265. 8./. tevor
235 rebicte @#6 target 235 .ovrce% a. 24 O*% ? 16/@/6 4,/ 8./ #@ 426=/4 235 .#867/ 8./5 ;:
?3586B,:2C .vra 3% (XC 24 (WA(X] 235 J/#6=/ e2B#@@ ` M26B U#,3.#3C More 1bav Coot Rea.ov: .
ieta Cviae to Poetic Metabor \> Q[% #@ Z,-% K6/.. (O>OR%
lall 2004| !"#$%&'( #'* +,&'( -'#./(0 &' 1#2

15
underlying idea or principal subject. 1he source, which carries the metaphor,
is the domain rom which the salient eatures are drawn and then attributed to
the target. So, in the simple metaphor lawyers are pigs,` lawyers` is the
target and pigs` is the source. 1he source imports a host o eatures
associated with pigs` without seeking to draw an identity-relation between
lawyers and pigs. 1he associations here might include rapacious appetite,`
non-discriminating appetite,` eating at the trough,` growing at,` ailing
to do any work,` and een perhaps absence o personal hygiene.` Analogies
work in the same way: eatures rom the source are mapped onto the target,
importing a series o elements that are not present in the target.
One o the irst studies o analogical mapping between domains was Gick
and lolyoak`s reinterpretation
22
o Duncker`s early radiation problem.
23
In
Duncker`s experiment, subjects were asked to sole the problem o curing a
cancerous tumor by radiation therapy. 1he constraints on any solution were:

ligh energy radiation would damage the patient`s tissues.
Low energy rays would not kill the tumor.

1he solution lay in directing a number o low energy rays rom dierent
positions on the body, but which intersected at the tumor site and created a
point o high energy. In this way the two constraints were satisied. lew
subjects soled the problem when presented in this way.
Gick and lolyoak
24
adapted this problem to inestigate analogical
reasoning. 1heir new experiment included a number o stories that the
subjects read prior to tackling the radiation problem. One story inoled an
army o men which sought to oerthrow a tyrannical ruler who was hiding in
a ortress. Unortunately or the rebels, the roads leading to the ortress were
all mined, such that i a large number o men walked oer any one road a
mine would explode. 1he solution was to break the men into small groups
and send them down dierent roads to conerge on the ortress at the same
time. 1he analogy between the ortress and the tumor was now obious, and
the number o subjects soling the problem rose dramatically, especially when
the subjects were prompted that the solution to the radiation problem might
lie in one o the preious stories. 1his demonstrated the power o analogical
mapping rom the source ,tyrant story, to the target ,cancer story,, where the
structural elements were the same and where the outcome o the source story
was the same as the desired one or the target.
1hough research has shown that analogical reasoning inoles some kind
o mapping between domains, this is only the start o any model o analogy.

22. M.L. Gick & Keith J. lolyoak, -'#./(&$#. 34/5."6 7/.8&'(, 12 Cognitie Psychol. 306
,1980, |hereinater Gick & lolyoak, -'#./(&$#. 34/5."6 7/.8&'(|, M.L. Gick & Keith J. lolyoak,
7$%"6# 9'*:$;&/' #'* -'#./(&$#. !4#',<"4, 15 Cognitie Psychol. 1 ,1983, |hereinater Gick &
lolyoak, 7$%"6# 9'*:$;&/'|.
23. Karl Duncker, =' 34/5."6>7/.8&'( ,Lynne S. Lees trans., Am. Psychol. Assn. 1945,.
24. Gick & lolyoak, -'#./(&$#. 34/5."6 7/.8&'(? ,:@4# n. 22, at 306, Gick & lolyoak,
7$%"6# 9'*:$;&/', ,:@4# n. 22, at 1.
!"#$%&' ") *+, -.."/0&*0"% ") 1,2&' 3$0*0%2 40$,/*"$. |Vol. 2

158
lirst, the question arises as to 5+&* is mapped rom source to target. 1he
mapping seems to inole some structural elements rather than merely surace
eatures. 1hus, in the radiation problem, the important concepts seem to be
the structural concepts o attack ,by ray or soldiers,, the concept o splitting
the attack ,ray or soldiers,, and the concept o the attacked item being
malignant ,cancer or tyrant,. 1he surace eatures can be ,and indeed must
be, ignored: one story inoled the body o a patient and the other inoled a
ictitious country, one story inoled soldiers and the other radiation rays, and
so orth.
Notwithstanding that structure seems to be the key to mapping here, it
cannot be the entire solution to the question o how analogy operates.
Research into the retrieal o analogs indicates that humans ind retrieal o
structural analogs ery diicult, and ind analogs based on surace eatures
much easier to recall.
25
In studies on the radiation problem, subjects more
oten recalled an analog about a doctor using rays or some non-therapeutic
reason than they recalled the tyrant analog, een though the doctor story was
not helpul in soling the problem. Structural isomorphism, or the need to
see structural similarities between the source and the target is an important
constraint upon analogical inerence, but other constraints exist, as I shall
shortly explain.
1he second question is 5+6 certain eatures are mapped. In another
amous example,
26
students learned more readily about the mechanics o the
atom by using the analogy o the solar system. In this analogy, the student
mapped the known concept o the sun being at the heart o the solar system,
onto the unknown world o the atom. 1hus, the atom`s nucleus became the
sun,` and the electrons became the circling planets.` 1he known concept
o attraction between sun and planets was used as an analog or the attraction
between nucleus and electrons. Both o the eatures mapped are structural
eatures o the source. loweer, other equally plausible structural eatures o
the source are not mapped onto the target. 1hus, the inormation that the
third planet rom the sun has lie on it is not mapped onto the third electron
in an atom, the act that the sun is largely gaseous is not mapped onto the
nucleus, and so on. 1o ind out why some eatures are chosen while others
are ignored, we need some account o what constraints inluence the mapping
procedure. 1his leads to the multiple-constraint model.


25. Gick & lolyoak, -%&'"20/&' 7$"8',9 :"';0%2< .#=$& n. 22, at 306, Dedre Gentner et al.,
>+, ?"', ") :090'&$0*6 0% >$&%.),$, 25 Cognitie Psychol. 431 ,1993,, Keith J. lolyoak & K. Koh,
:#$)&/, &%@ :*$#/*#$&' :090'&$0*6 0% -%&'"20/&' >$&%.),$, 15 Memory and Cognition 332 ,198,, Mark
1. Keane, A% ?,*$0,;0%2 -%&'"2#,. 3+,% :"';0%2 7$"8',9., 39A Quarterly J. o Lxperimental
Psychol. 29 ,198,. B). C.M. \harton et al., C,'"5 *+, :#$)&/,D -%&'"20/&' :090'&$0*6 &%@ ?,*$0,;&'
B"9=,*0*0"% 0% ?,90%@0%2, 14 Cognitie Psychol. 246 ,1994,.
26. Dedre Gentner, :*$#/*#$, E&==0%2D - >+,"$,*0/&' F$&9,5"$G )"$ -%&'"26, Cognitie Sci.
155 ,1983,, Dedre Gentner & D.R. Gentner, F'"50%2 3&*,$ "$ >,,90%2 B$"5@.D E,%*&' E"@,'. ")
H',/*$0/0*6, in E,%*&' E"@,'. 99 ,Dedre Gentner & A.L. Steens eds., Lawrence Lrlbaum Assocs.
1983,.
lall 2004| !"#$%&'( #'* +,&'( -'#./(0 &' 1#2

159
B. 1he multiple-constraint model

Keith lolyoak and Paul 1hagard
2
hae presented a multi-constraint`
model o analogy. 1hey suggest that arious constraints inluence the
generation and perception o analogies. 1hey suggest three basic kinds o
constraints in analogical thinking:
28
constraints at the surace leel, at the
structural leel, and at the purposie leel. In the sections that ollow I
outline these leels and demonstrate how they apply to analogies in law.

34 !%" ,567#$" $/',86#&'8

1he surace-leel constraint suggests that an analogy is guided by the *&6"$8
,&9&.#6&80 in the surace-leel elements in the source and target domains. 1hus,
in the example o the atom-solar system analogy, there are directly similar
eatures present in both domains. In each domain, something lies in the
center o the system, and other things reole around this central body. 1hus,
it is immediately possible to establish a one-to-one mapping between surace-
leel objects within the two domains.
29

low does this apply to law, and in particular, to the erry example
Surace-leel similarities are obious between the erry case and the train: both
inole traelers moing between two places. Alternatiely, the similarities
between the erry and the hotel cases are obious: both cases inole locked
rooms ,whether hotel room or erry stateroom, rom which the bags were
taken, both hae restaurants, bars, and so on. Surace similarity exists in both
cases.
laing suggested that the surace-leel similarity constraint might operate
in the erry example, the question is whether there is stronger eidence that it
is maniested in legal reasoning. 1here are two strong pieces o eidence or
this. lor a start, there is the obious application o this constraint when a
judge assesses two cases as analogous based on surace-leel eatures. Say, or
example, we are considering lower-court decisions on the distribution o
property between spouses ater diorce. Let us say a precedent is similar to
the current case - in both cases there are three children o the marriage, the
wie has custody o them, the major asset is the house, the wie sued or
diorce based on the husband`s adultery, and so orth. It is likely that the
judge will, without examining any deeper structural eatures or principles,

2. Keith J. lolyoak & Paul 1hagard, - :/9;58#8&/'#. </*". /7 -'#./(&$#. =6/>."9 ?/.@&'(,
in ?&9&.#6&80 #'* -'#./(&$#. A"#,/'&'( 242 ,Stella Vosniadou & Andrew Ortony eds., Cambridge U.
Press 1989,, lolyoak & 1hagard, ,5;6# n. , Paul 1hagard, B&9"',&/', /7 -'#./(0, in -'#./(&$#.
A"#,/'&'(, ,5;6# n. 4 at 105.
28. lolyoak & 1hagard, ,5;6# n. , at 5.
29. 1his type o similarity matching is oten called propositional` similarity because
matching occurs between propositions rather than between the relations that operate on the
domains. ?"" &*4 at 24-31.
]ovrvat of tbe ...ociatiov of egat !ritivg Director. |Vol. 2

160
award a property diision based on the surace-leel similarity in the cases.
30

1his is more likely i, as commonly happens, the judge is aced with many
cases to decide in a limited time rame. \e should expect to see broadly
similar results in the sentencing o conicted criminals, the assessment o
liability in automobile accidents, and other areas where lower-leel courts are
dealing with run-o-the-mill actions.
31

Surace-leel similarity matching has also been demonstrated by Alan
1yree. 1yree reported on a study o the decision making o law students in
the ield o troer, that is, the property law domain dealing with ownership o
lost chattels.
32
le asked upper-leel law students to predict the outcome o
an undecided case based on a set o appellate decisions.
33
le showed that law
students` judgments o similarity were not based on high-leel analysis o
policy considerations or on legal doctrine, but rather on matching surace-
leel acts.
34
le was then able to use this inormation to build a simple
computer-based expert system that accurately predicted the outcome o cases
based on simple act matching.
35
James Popple later adopted this approach
and showed that it was a generalizable methodology in a number o dierent
legal domains, including copyright and employment law.
36
Popple also built
an expert system based on this methodology, and showed that it had a high
degree o accuracy in predicting the outcome o cases. 1hese two studies in
law, together with the many other cognitie psychological studies, gie strong
initial indications that surace-leel mapping is almost certainly undertaken in
precedential reasoning and legal analogy-making generally.
3


30. Knowledge engineering exercises hae shown this occurs in low leel domains such
as diorce proceedings. ee John Zeleznikow, Andrew Stranieri & Mark Gawler, tit|: .
egat ert ,.tev !bicb Detervive. Proert, Diri.iov |ov Dirorce, 3 Artiicial Intelligence and L.
26-25 ,1996,.
31. lor examination o these types o commonplace cases and the role that they play in
knowledge engineering, see Dan lunter, ^ear Kvorteage: vavctire earvivg ,.tev. iv ar, 5 Va.
J.L. & 1ech. 9 ,2000,.
32. Alan L. 1yree, act Covtevt .vat,.i. of Ca.e ar: Metboa. ava ivitatiov., 22,1,
Jurimetrics J. 1 ,1981,.
33. 1hese cases were all drawn rom the Lnglish louse o Lords, and ligh Court o
Appeals. ee Alan L. 1yree, ert ,.tev. iv ar 161-5 ,Prentice lall 1989,, 1yree, .vra n.
32.
34. Amos 1ersky, eatvre. of ivitarit,, 84 Psychol. Re. 32 ,19,, 1yree, .vra n. 32.
35. Alan L. 1yree, ivaer: .v ert ,.tev, in 1be Proceeaivg. of tbe ortietb .vvvat
Covferevce of tbe .v.trata.iav |virer.itie. ar cboot. ...ociatiov ,AULSA 1985,, 1yree, .vra n. 33.
36. James Popple, b,.ter: . Pragvatic egat ert ,.tev ,unpublished Ph.D. dissertation,
Dept. o Computer Sci., Australian Natl. U. 1993, ,copy on ile with author,, James Popple, .
Pragvatic egat ert ,.tev ,Dartmouth 1996,.
3. 1hese two studies hae signiicant problems or my purposes: they are based only on
law student prediction and not actual adjudication. 1here is also the concern about what
accurate predictions o outcomes` means and whether it is a good basis or a theory o
analogy in law. loweer, these concerns can be answered by suggesting that the process o
student reasoning is consistent with adanced legal reasoning by judges and attorneys. ee Gary
L. Blasi, !bat ar,er. Kvor: ar,erivg erti.e, Cogvitire cievce, ava tbe vvctiov of 1beor,, 45 J.
Leg. Lduc. 313 ,1995,.
lall 2004| !"#$%&'( #'* +,&'( -'#./(0 &' 1#2

161
Moreoer, surace-leel similarity depends on context. lolyoak and
1hagard note that similarity is not a ixed concept, especially surace-leel
similarity. 1he perception o similarity diers depending on the background
inormation which is proided, the context eect.` In classic studies
undertaken by 1ersky,
38
subjects were asked which one o a group o
countries was most similar to Austria. \hen asked which country was most
similar to Austria rom a group including Sweden, lungary, and Poland,
subjects generally chose Sweden. loweer, when presented with the group o
Sweden, lungary, Norway and asked which was most similar to Austria,
subjects usually chose lungary. Keeping two o the possibilities constant,
Sweden and lungary, but changing one possibility, Poland swapped or
Norway, was enough to change the assessment o similarity een though the
changed country was not deemed to be the closest.
1he explanation is simple: in the irst test, the similarities between Poland
and lungary ,at the time both were Lastern Bloc communist under Soiet
control, grouped them together, leaing Sweden to appear more like Austria
- emphasizing their similarity in terms o their being \estern Luropean in
orientation and capitalist. In the second test, the similarities between Sweden
and Norway ,similar language and both being Scandinaian,Nordic countries,
grouped them together, leaing Austria to appear more like its old imperial
ally, lungary. 1hus, human assessment o similarity on surace eatures has
been shown to be context-dependent.
Do we see similar context dependence in law 1here are three pieces o
eidence airming this. lirst, Kelman, Rottenstreich, and 1ersky undertook
a study o the eect o context dependence in legal decision making.
39

1hough they were looking or slightly dierent eects rom the one posited
here, they demonstrated that people generally displayed the same context
eects when making legally oriented decisions. 1hey concluded that whether
intended or not, adding alternaties will not only introduce what may seem a
substantiely plausible decision option but will alter the choice patterns
among other options.`
40

1his doesn`t explain eery part o the analogy-making process in law,
since these studies did not examine judges` decision making, nor were they
looking at analogical reasoning. loweer, their results are extremely
suggestie. Indeed it would be remarkable i context eects ound in one
type o legal reasoning were not also present in precedential reasoning.
lollowing on rom this, I suggest that the process o adocacy beore a
judge is, in part, an attempt to introduce this context eect. Recall the erry
example. Counsel or Plainti, seeking to adance the loating hotel` theory,
with its attendant liability outcome, will presumably highlight the actual

38. 1ersky, ,345# n. 34. 1his has been replicated in many dierent studies, and in
dierent ields.
39. Kelman, Rottenstreich, and 1ersky, 6/'7"879:"4"'*"'$" &' 1"(#. :"$&,&/' ;#<&'(, 25 J.
Leg. Stud. 28 ,1996,.
40. =*> at 310.
!"#$%&' ") *+, -.."/0&*0"% ") 1,2&' 3$0*0%2 40$,/*"$. |Vol. 2

162
similarities between the erry and a hotel and may introduce other examples
that make the seagoing train` theory less palatable. 1hus, a good lawyer
might introduce another example where a metropolitan trolley-car company
was held not liable or thet committed on its trolleys.
41
1hus, the context
eect experiment becomes: Out o ,hotel, train, trolley car, which is a erry
more like` 1he answer is, presumably, the hotel. Deendant`s counsel might
counter by introducing another example, where the owners o a permanently
berthed cruise-liner were ound liable or water damage to a guest`s luggage.
1he context eect experiment then becomes: Out o ,hotel, berthed liner,
train, which is a erry more like` In introducing these new examples, the
attorney endeaors to orce a new classiication o similarity on the problem,
in the same way that introducing Poland` or Norway` aects our iew o
Austria. 1he best trial lawyers are able to inluence judicial assessments o
similarity by a skillul manipulation o the context eect.
linally, the context eect explains an obseration made in the early work
on legal reasoning by Ldward Lei. In his seminal study, Lei charted the
deelopment o the doctrine o inherently dangerous goods.
42
It began with
the single decision that an exploding lamp was dangerous, eentually
expanded to included things imminently` dangerous i deectie, and
eentually became things inherently` dangerous. le concluded:

|1he concept o inherently dangerous goods| . . . has the capacity to
suggest by the implication o hypothetical cases which it carries and
een by its ability to suggest other categories which sound the same.
1he phrase imminent danger,` or example, suggested immediacy,
inherence, and eminence. 1o this extent, the phrase suggests the
instances to be included under it . . . .
43


1he context eect o similarity seems to be operating here, though this time
the eatures making up the context are not other cases but the words o the
doctrine. Imminent` suggests immediacy` and inherent` and thus the
context aects the articles that subsequently are included in the category. I do
not present this eidence as incontroertible, as no studies on this point in law
hae been made. loweer, it seems, like the eidence aboe, to point to the
same eect in legal analogical inerence.
As a inal obseration on surace-leel similarity, it is important to bear in
mind that this type o similarity is only an initial constraint and can oten be
satisied by many things. lor example, as an analog o the atom, we could
suggest a ball on a string whirled aboe our heads or cars circling the outside

41. It should not matter whether this is an actual precedent or merely an example
proided by counsel. Note also that this is introducing a notional point or the other side` by
proiding an example that reutes liability. It is also possible to introduce examples that point
to liability, but this is a less interesting eect than the one described here.
42. Lei, .#5$& n. 5, at 9-2.
43. 67. at 2.
lall 2004| !"#$%&'( #'* +,&'( -'#./(0 &' 1#2

163
lane at Dupont Circle in \ashington D.C. lor the constraint o surace-leel
similarity, these analogs are almost as good as the solar-system analog.
loweer, we immediately recognize that the 3#..4/'4#4,56&'( analog and the $#64
&'4789/'54:&6$." analog are somehow less compelling or useul than the solar
system analogy, so some other constraints must be operating here. Lqually in
the legal erry example, many other cases with surace similarity may exist,
none o which make good analogies. \hat o cases where a passenger on a
cruise-liner is killed by another passenger or where a hotel-keeper has
discriminated against emale sta 1hese are precedents which contain
certain surace-leel similarities to the target ,undecided, case, but which are
obiously poor analogies. \e can see thereore that this type o similarity is
not the complete answer, other constraints must be operating to inluence our
assessment o whether the analogy is a ery good one.

;< !%" ,568$586" $/',56#&'5

lolyoak and 1hagard`s second constraint is ,568$586". 1his constraint
inoles the pressure to identiy consistent structural parallels between the
two domains. Consider the solar-system example again. \e hae the
ollowing mappings at the surace leel:

Source: Solar system maps 1arget: Atom
Sun Nucleus
Planets Llectrons
lAS_VLLOCI1\
,planet,
lAS_VLLOCI1\
,electron,
1LNDS_1O_lL\_OU1
,planet,
1LNDS_1O_lL\_OU1
,electron,

1he surace leel is sometimes called the propositional leel because it
operates on the propositions that describe the source.
44
lAS_VLLOCI1\
,x, is a proposition describing an aspect o the solar system. Propositions are
predicates that hae only one slot. lor example, the predicate
lAS_VLLOCI1\ ,x, is a proposition indicating that a body, x, has a elocity,
the slot x can be illed with planet, electron, Boeing_4, or
President_Bush_in_a_calacade_driing_down_Connecticut_Aenue. 1hus,
there is similarity at propositional leel o these our things: they all hae
elocity.
In the solar system-atom analogy, there is a basic mapping between the
objects at propositional leel. loweer at the structural leel, where we
describe the relationships between objects, similarity is more obious:

44. lolyoak & 1hagard, ,896# n. , at 24-31.
!"#$%&' ") *+, -.."/0&*0"% ") 1,2&' 3$0*0%2 40$,/*"$. |Vol. 2

164

Source: Solar system maps 1arget: Atom
A11RAC1S
,sun, planet,
A11RAC1S
,nucleus, electron,
ORBI1S
,planet, sun,
ORBI1S
,electron, nucleus,

1his is called the relational leel. A11RAC1S ,x, y, is a basic relation o
the system.
45
Relations are predicates that hae more than one slot. 1hey are
called relations` because they relate one or more o the slot-illers to the
other or others. So, the predicates IS_PAR1_Ol ,a, b, or ORBI1S ,x, y, are
relations, indicating that an entity a is part o b, or that an entity x orbits y,
respectiely. Structural mapping inoles a mapping o relations, not
propositions.
1here is a strong structural mapping in the atom-solar system analogy in
that: ,1, the outer bodies hae a elocity, ,2, and a tendency to moe
outwards, ,3, but the attraction pulls them inwards, ,4, and as a consequence
o this the outer bodies orbit the inner one, and so orth. 1hus, there is close
structural similarity between the relations o the objects. 1his is not the case
with the analogy o the car circling Dupont Circle: apart rom the surace leel
similarity, there is little in the relations that map rom the source to the target.
Perhaps we could suggest that the relation ORBI1S is the same, but een this
is stretching the natural meaning o orbits.`
\ithin law we see the structure constraint gie rise to an important
characteristics o legal reasoning. Relational similarity typically occurs when
we compare legal principles or concepts in two dierent cases. In the
example o the erry and the missing bags, we can draw an analogy with the
hotel-keeper precedent by reerence to the isomorphism at the relational leel
as well as at the propositional leel shown here:

Source: lotel Precedent Maps to 1arget: lerry Case
liable_in_bailment
,hotel_keeper, guest,
liable_in_bailment
,erry_line, passenger,
owes_duty_to
,hotel_keeper, guest,
owes_duty_to
,erry_line, passenger,
standard_breached
,hotel_keeper,
standard_breached
,erry_line,
sleeps_oer ,guest, sleeps_oer ,passenger,
has_restaurant ,hotel, has_restaurant ,erry,
missing ,bags, missing ,bags,
Ltc. Ltc.


45. 567
lall 2004| !"#$%&'( #'* +,&'( -'#./(0 &' 1#2

165
Alternatiely, we could suggest the analogy to the railroad by reerence to
the dierent relational structure that is possible to project onto the erry case.
1he adequacy o the analogy is dependent on our ability to it the target
within the relational structure o the source. loweer, the constraint is not
absolute, or it is possible to project a number o mappings onto the target:
either o the loating hotel` or seagoing train` analogies is structurally as
coherent as the other. 1his seems to accord with the eeling that in the erry
case both analogies are good ones, and the outcome could go either way.
1his gies rise to the question o why we ocus on the structural aspects
we do, and not on an ininity o other structural eatures. Answering this
question will explain why we ind the solar system analog better than the car-
in-Dupont-Circle analog. Lqually, we need an explanation why other
relational eatures between the solar system and the atom are not presered,
such as distance between planets or the relatie sizes o planets and the sun.
Or within the legal example, why we chose the eatures explained aboe,
rather than any other potentially releant eature o other precedents.
One answer to this problem is that the other constraints - surace and
purpose leel - operate in conjunction with the structure constraint. But
within the structure constraint itsel, there is also another mechanism that
aects the choice o analogs. Dedre Gentner`s model o the structural
constraint suggests that relational mapping at the structural leel relies on the
principle o systematicity.` 1his is the idea that people preer to map
systems o predicates that contain higher-order relations with inerential
import rather than to map isolated predicates. 1hat is, we pick clustered
groups o relations which are able to explain why the system works as it does,
rather than pick isolated predicates which may be similar but which are
singletons and do not help to explain the system. In this, she expresses what
she calls our tacit preerence or coherence and deductie power in
interpreting analogy.`
46
In the solar system-atom analogy, the relations which
explain why the system operates as it does are the ones which are mapped.
1hus, the relation A11RAC1S in both the source and target explains why the
outer bodies ,haing a VLLOCI1\, end up ORBI1ing the inner body.
1his principle o systematicity is important in legal analogies. 1his
principle encodes the human predilection or choosing structural mappings
which are higher-order and which are explanatory o the entire system. \e
should thereore expect to see certain analogies preerred i they operate at a
higher, causatie leel than the alternaties. 1ake our erry example again, but
change the precedents slightly. Let us assume that the reason the train
precedent was decided against the plainti was not because o bailment law,
but rather because o some other policy limiting liability o state-run
industries.
4
1he highest-leel, causatie relation in the erry case would
probably be LIABLL_IN_BAILMLN1_LA\ ,erry_line, passenger, But in
the train example, the highest relation is NO1_LIABLL_IN_PUBLIC_LA\

46. Gentner et al.3 ,456# n. 4, at 12.
4. In the days when railroad companies were oten or usually a state-run industry.
!"#$%&' ") *+, -.."/0&*0"% ") 1,2&' 3$0*0%2 40$,/*"$. !"#$% '

())
*+,-$.,/0 1,223453+6 #+ 2#73 2-7-$,+ +3$,8-#4% 9#8.-8:28,4;-45 8:3
1+#1#2-8-#4,$ $3<3$ ,4; $#.3+=$3<3$ +3$,8-#4,$ 2-7-$,+-8-320 ,8 8:3 :-5:328 $3<3$
8:3+3 -2 , 7-27,8>: ?38.334 8:3 @3++/ >,23 ,4; 8:3 43. +,-$.,/ 1+3>3;348% A3
,+3 8:3+3@#+3 $322 $-B3$/ 8# @-4; 8:-2 43. +,-$.,/ 1+3>3;348 ,2 C23@C$ ,4 ,4,$#5/
,2 8:3 #$; 1+3>3;348%

56 7+, 8#$8"., /"%.*$&0%*

D#$/#,B ,4; E:,5,+;F2 8:-+; >#428+,-48 #4 8:3 ,4,$#5/ -2 8:3 8#$8"., @#+
>#42-;3+-45 8:3 ,4,$#5/ ,8 ,$$% G4 ,;#18-45 8:3 2#$,+ 2/2837=,8#7 ,4,$#5/0 .3
:,<3 , 4C7?3+ #@ 1C+1#2320 -4>$C;-45 8+/-45 8# C4;3+28,4; .:/ 8:3 ,8#7
+37,-42 -48,>80 +,8:3+ 8:,4 >#$$,12-45 -4.,+;2 #4 -823$@ #+ @$/-45 #C8.,+;2 -4 ,$$
;-+3>8-#42% E:3 2#$,+ 2/2837 ,4,$#5/ 1+#<-;32 8:3 ?,2-2 @#+ 8:-2
C4;3+28,4;-45% E:3 1$,4382 .#C$; @$/ #C8 8# 8:3 @,+ +3,>:32 #@ 8:3 5,$,H/ .3+3
-8 4#8 @#+ 8:3 IEEJIKE-45 3@@3>8 #@ 5+,<-8/
D#.3<3+0 8:-2 ;#32 4#8 3H1$,-4 #8:3+ @3,8C+32 #@ 8:3 ,8#70 2C>: ,2 8:3
4,8C+3 #@ 8:3 4C>$3C2 #+ >3+8,-4 LC,48C7 73>:,4->,$ @3,8C+32 #@ 8:3 ,8#7% G4
#+;3+ 8# C4;3+28,4; 8:323 @3,8C+320 8:3 2#$,+ 2/2837 ,4,$#5/ -2 C23$3220 ,4; .3
7C28 @-4; #8:3+ ,4,$#5-32 8# 3H1$,-4 8:3 1+#>322% M#+ 3H,71$30 8:3+3 -2 ,4
,4,$#5/ #@ 8:3 ,8#7 8# , N>$#C; #@ 3$3>8+#420O .:->: -2 C23@C$ 8# 3H1$,-4 2#73
3$3>8+#=<,$34>/ >#4>31820 ,4; .:->: -2 >3+8,-4$/ ?3883+ 8:,4 8:3 2#$,+=2/2837
,4,$#5/ @#+ 8:-2 1C+1#23%
A-8:-4 $,.0 8:3 1C+1#23 >#428+,-48 >$3,+$/ ,11$-32% E,B3 8:3 @3++/ >,23 ,2
,4 3H,71$3% E:3 1C+1#23 #@ 8:3 ,88#+43/ @#+ P$,-48-@@ -2 8# 7,B3 8:3 7#28
1$,C2-?$3 >,23 @#+ P$,-48-@@% D34>3 ,4 ,4,$#5/ 8:,8 #13+,832 ,5,-428 P$,-48-@@F2
-483+328 Q 2C>: ,2 8:3 8+,-4 ,4,$#5/ Q -2 >$3,+$/ $322 ;32-+,?$3 ,4; $322
>#713$$-45 8:,4 8:3 ,$83+4,8-<3 :#83$ ,4,$#5/% E:3 #11#2-83 -2 8+C3 @#+
R3@34;,48F2 >#C423$% E:C20 8:3 1C+1#23 @#+ .:->: $,./3+2 .-2: 8# C23 ,4
,4,$#5/ .-$$ -4@$C34>3 8:3-+ 13+>318-#4 #@ -8S ,4; ,2 , >#423LC34>3 8:3 1C+1#23
>#428+,-48 .-$$ ;383+7-43 8:3 ,4,$#5/ -2 , 1,+8->C$,+$/ 5##; #43 #+ ?,; #43 @#+
,4/ 5-<34 1,+8/%
E:3 1C+1#23 >#428+,-48 ,$2# ,11$-32 8# TC;->-,$ ,;TC;->,8-#4% U-4>3 8:3
-48+#;C>8-#4 #@ I73+->,4 V35,$ J3,$-27
WX
,4; K+-8->,$ V35,$ U8C;-32
WY
-8 :,2

WX% I73+->,4 $35,$ +3,$-27 ?35,4 -4 8:3 $,83 4-4383348: ,4; 3,+$/ 8.348-38: >348C+-32 ,2
,4 ,88,>B #4 @#+7,$-270 .:->: C48-$ 8:340 .,2 8:3 1+3;#7-4,48 ,11+#,>: #@ TC;->-,$ 8:#C5:8% G4
2#73 23423 -8 .,2 , +3,>8-#4 8# 8:3 3@@#+82 #@ $35,$ 8:3#+-282 $-B3 K:+-28#1:3+ K#$C7?C2 V,45;3$$
#+ J#2>#3 P#C4; 8# 7,B3 $,. 7#+3 $-B3 ,4 ,?28+,>8 2>-34>3% 9,, Z,+5,+38 R,<-320 -.:0%2 *+,
1&; <#,.*0"% YW=((Y *U.338 [ Z,H.3$$ (YYW6% E:3 I73+->,4 $35,$ +3,$-27 2>:##$ +31+323483; ,
@#+>3@C$ ,4; -4@$C348-,$ +3,>8-#4 ,5,-428 8:-2 8+34;0 ,4; -4;33; ,5,-428 ,$$ 8:3 $35,$ 8:3#+-32 .:->:
371:,2-\3; 8:3 ,4,$/8->,$ ,4; ,?28+,>8 #<3+ 8:3 ;32>+-18-<3 ,4; 8:3 >#4>+383% 9,, D-$,-+3
Z>K#C?+3/ [ 9-53$ R% A:-830 7,=*>"": "% !#$0.8$#?,%/, (X]=XX *^$,>B28#43 P+322 (YY_6S `%a%
J-;;,$$0 !#$0.8$#?,%/, (b_=bW *^C883+.#+8:2 (YY(6S `%A% D,++-20 1,2&' @+0'"."8+0,. *^C883+.#+8:2
(YXc6%
WY% E:3 7#<37348 ?35,4 ,8 Z,;-2#40 A-2>#42-40 -4 (Y]] ,8 8:3 1#+8348#C2$/ 4,73;
NK#4@3+34>3 #4 K+-8->,$ V35,$ U8C;-32%O 9,, `%D% U>:$353$0 A"*,. 7";&$?. &% B%*0C&*,D E80%0"%&*,?D
&%? -)),/*0"%&*, F0.*"$G ") *+, H"%),$,%/, "% H$0*0/&' 1,2&' 9*#?0,.D _) U8,4% V% J3<% _Y( *(YXW6S
lall 2004| !"#$%&'( #'* +,&'( -'#./(0 &' 1#2

16
become a commonplace to suggest that laws do not completely determine
legal outcomes and that judges hae considerable leeway in their decision
making. Justice Olier \endell lolmes amously remarked:

General propositions do not decide concrete cases . . . . I always say .
. . that no case can be settled by general propositions, that I will
admit any general proposition you like and decide the case either
way.
50


1his is oten characterized as an example o lolmes`s rule skepticism:
51
that
is, the notion that legal rules do not entail particular legal results.
52
loweer it
is not hard to see this as an example o the analogical purpose constraint,
working within judicial decision making.
53

More recently, we hae seen the Critical Legal Studies moement
introduce related, but more politically charged, approaches to legal reasoning.
lor example, Duncan Kennedy examined how he might decide a case in
which the precedents were arrayed against the outcome he preerred.
54
le
discussed the approach o a judge by examining the internal conersation a
judge has with hersel. lis conclusion about legal reasoning proides a clear
example o the purpose constraint in operation:
Legal reasoning is a kind o work with a purpose, and here ,in
adjudication, the purpose is to make the case come out the way my sense o
justice tells me it ought to, in spite o what seems at irst like the resistance or
opposition o the law.`
55

\e can see the purpose constraint operating in legal settings when an
attorney is presenting an argument in aor o her client, but equally we see it
in theories o adjudication such as those presented by the Realists and the
proponents o Critical Legal Studies.


Roberto Mangabeira Unger, !%" 34&5&$#. 1"(#. 657*&", 8/9":"'5 ,larard U. Press 1986,, Roberto
Mangabeira Unger, !%" 34&5&$#. 1"(#. 657*&", 8/9":"'5, 96 lar. L. Re. 561 ,1986,.
50. 6"" \. L. Rumble, -:"4&$#' 1"(#. ;"#.&,:< 6=">5&$&,:? ;"@/4: #'* 5%" A7*&$&#. B4/$",, 39-
40 ,Cornell U. Press 1968,.
51. Jerome lrank, 1#2 #'* 5%" 8/*"4' 8&'* xi ,Doubleday & Co. 1963,.
52. 3@. the position expressed by a noted Realist, who was also a contemporary o
lolmes. Karl N. Llewellyn, !%" 3/::/' 1#2 !4#*&5&/'< C"$&*&'( ->>"#., 38 ,Little, Brown 1960,
,|R|ules which make sense on their ace, and which can be understood and reasonably well
applied een by mediocre men . . . hae a air chance to get the same results out o dierent
judges, and so in truth hit close to the ancient target o a goernment o laws and not men.`,.
53. Lqually, the act skepticism championed by Jerome lrank has certain eatures in
common with the analogical purpose constraint. 6"" lrank, ,7>4# n. 51, at xiii.
54. Duncan Kennedy, D4""*/: #'* 3/',54#&'5 &' -*E7*&$#5&/'< - 34&5&$#. B%"'/:"'/./(0, 36 J.
Leg. Lduc. 518 ,1986,.
55. F*. at 526 ,emphasis remoed,.
!"#$%&' ") *+, -.."/0&*0"% ") 1,2&' 3$0*0%2 40$,/*"$. !"#$% '

()*
!!!" $%&'()*+%&

+,-./ 01. 2#3/,0,-. 42,./2. 522#6/0 #7 5/5$#389 5/: 01. .-,:./2.
;<.4./0.: 5=#-. 0150 01. >6$0,;$. 2#/40<5,/0 >#:.$ 5;;$,.4 ,/ $5?9 ?150
$.44#/4 >,310 ?. 5;;$8 0# 01. 0.521,/3 #7 5/5$#38 ,/ $5? 421##$4 5/: 01. 64.
#7 5/5$#38 =8 ;<520,2,/3 $5?8.<4@
A,<409 B 4633.40 0150 01,4 >#:.$ ;<#-,:.4 5 >#<. :.05,$.: ;,206<. #7 ?150 ,4
15;;./,/3 ?1./ $5?8.<4 2#/40<620 5/5$#3,.4 5/: #7 ?150 C6:3.4 5<. :#,/3
?1./ 01.8 5:#;0 5 ;5<0,26$5< ;<.2.:./0% D06:./04 5/: $5?8.<4 25/ =.00.<
2<,0,E6. 254.49 54 ?.$$ 54 ,/:,-,:65$ 5/5$#3,.49 ,7 01.8 6/:.<405/: 0150 01.<. 5<.
>6$0,;$. $.-.$4 50 ?1,21 5/ 5/5$#38 25/ =. 2#/40<620.: 5/: 5/5$8F.:%
D.2#/:9 01. >#:.$ ;<#-,:.4 5 7<5>.?#<G 0# 3./.<50. ,/0.<.40,/3 5/: >#<.
2#>;.$$,/3 5/5$#3,.4 =8 406:./04 5/: $5?8.<4% H.3,//,/3 $5? 406:./04 5/:
$.44 .I;.<,./2.: $5?8.<4 0./: 0# 2#/40<620 5/5$#3,.4 0150 5<. =54.: #/ 46<752.J
$.-.$ 4,>,$5<,0,.4% K4 01.8 =.2#>. >#<. ;<520,2.:9 =#01 405<0 0# <.2#3/,F. 0150
01. >#40 ;#?.<76$ 5/5$#3,.4 #;.<50. 52<#44 >6$0,;$. $.-.$4 5/: 01.8 $.5</ 0#
<.$8 #/ +./0/.<L4 4840.>50,2,08 ;<,/2,;$.9 >#40 $,G.$8 ?,01#60 .-.< <.5$,F,/3
0150 4621 5 2#/2.;0 .I,404 5/: 154 =../ :#26>./0.: .$4.?1.<.% M.2#3/,F,/3
54 .5<$8 54 ;#44,=$. ,/ 01.,< 25<..<4 0150 01.8 /..: 0# 2#/40<620 01.4. 4#<04 #7
$58.<.: 5/5$#3,.4 25/ #/$8 4;..: 01.,< :.-.$#;>./0 0#?5<:4 =.2#>,/3 .I;.<0
64.<4 #7 5/5$#3,25$ <.54#/,/3%
A,/5$$89 /6>.<#64 7,/:,/34 #7 2#3/,0,-. 42,./2. 25/ =. 64.: 0# 3./.<50.
>#<. ;.<4654,-. 5/5$#3,.4 N5/: ,/:..: >#<. ;.<4654,-. 5<36>./04 ,/ 3./.<5$O%
P/. .I5>;$. #7 01,49 :,42644.: 5=#-.9 ,4 Q-.<4G8L4 2#/0.I0 .77.20% H8 25<.76$
>5/,;6$50,#/ #7 18;#01.0,25$ 4,0650,#/49 5 $5?8.< 25/ >5G. 5 3##: 5/5$#38
5;;.5< .-./ >#<. 2#>;.$$,/3%